I build AI applications and model deployment tools.
A compact single-image vision-language model deployment project. My engineering work includes:
- Local/offline folder inference with JSONL and standalone HTML reports.
- Processor keyword-handling fixes and image-token/visual-embedding validation.
- Regression tests and documented CPU smoke-test results.
Selected contribution: the deployment tooling and validation work was selected for the XERJ Corpus Hub, with attribution and a pinned source revision.
Repository · Weights and model card · XERJ corpus
The model and core implementation come from Tether AI Research / VisionPsy Nano. My contribution focuses on the deployment workflow and reliability; the released base weights are unchanged.
